External data lakes, visualized.
Your customer data already lives in a data lake such as Snowflake or Databricks. Salesforce Data 360 (formerly Data Cloud) can copy it, query it where it lives, or read its files directly. Here's what each option actually does.
Choose an option to see what's involved and how the data moves:
In use involved in this option · Faded not used
When to use each
You choose per data stream, not per data lake. One lake can feed a heavily used loyalty table through acceleration and a rarely used archive through query federation.
Find the most cost-effective approach
The rule of thumb: federation pays for every row, every time it's read. Copying pays for each row once, plus changes. So the question is how many times each row gets read, and how often it changes. Query federation also runs on the lake's compute, which is billed separately. Answer five questions about one table.
Credit rates change over time and vary by contract. The comparisons below are relative, not quotes. Consult your contract for current rates.
Credit consumption rates change. Consult your Salesforce contract for current rates. The relative costs here are based on Salesforce's published list rates as of 2026 (pricing example, zero copy lessons, decision guide). A starting point, not a quote: rates vary by contract, and the lake's compute is billed separately. Model both sides with real query volumes.
Run the numbers
Federation is billed on the rows each read pulls back, every time. Copying in is billed once, plus whatever changes. Put in how a dataset will actually be used and see which costs less, including caching, where the refresh type changes everything. Start from either example, a daily segment that reads one month of orders or a daily calculated insight that reads all of them, then make the numbers your own. More on this in Zero copy isn't zero cost.
Run your own numbers
Pick an example, then change any number to make it yours.
How the data is used (applies to every option)
If acceleration is on (only affects the middle result)
What each option costs (credits)
First year with acceleration off: 37,800 credits. On, with full refresh: 36,100,000. Acceleration off is more than 20x cheaper.
Federate with acceleration off. Over the first year that's about 37,800 credits, against about 136,000 credits to copy it in.
For illustrative purposes only. Verify all numbers independently before implementing a solution. All figures are credits, at Salesforce list rates as of October 2026: federation at 70 per million rows read; copying in and caching at 2,000 per million rows loaded (batch). Rates change and contracts vary. Federation also costs compute on your data lake for each query, not shown here. Charges for querying data inside Data 360 apply to every option and are left out.
Two more things worth knowing
Zero copy works both ways. Data 360 can also share unified profiles, segments, and insights back to the lake without copying them, so the data team works with the same results.
Harmonization protects you from lake changes. Whatever the option, the data maps onto Data 360's data model before segments or agents use it. If a team renames a column in the lake, you fix one mapping instead of every segment. See data harmonization, visualized →
Simplified for learning, and capabilities change often: check current documentation for your platform. An independent tool, not affiliated with Salesforce. More Data 360 tools